Downloaded By: [Princeton University] At: 19:20 15 March 2007
Combust. Sci. and Tech., 179: 663–694, 2007
Copyright Q Taylor & Francis Group, LLC
ISSN: 0010-2202 print/1563-521X online
DOI: 10.1080/00102200600713583
Spontaneous Ignition of Pressurized Releases
of Hydrogen and Natural Gas into Air
Frederick L. Dryer, Marcos Chaos, Zhenwei Zhao, Jeffrey N. Stein,
Jeffrey Y. Alpert, and Christopher J. Homer
Mechanical and Aerospace Engineering, Princeton University, Princeton, NJ
Abstract: This paper demonstrates the ‘‘spontaneous ignition’’ (autoignition=inflammation and sustained diffusive combustion) from sudden compressed
hydrogen releases that is not well documented in the present literature, for which
little fundamental explanation, discussion or research foundation exists, and
which is apparently not encompassed in recent formulations of safety codes
and standards for piping, storage, and use of high pressure compressed gas systems handling hydrogen. Accidental or intended, rapid failure of a pressure
boundary separating sufficiently compressed hydrogen from air can result in
multi-dimensional transient flows involving shock formation, reflection, and
interactions such that reactant mixtures are rapidly formed and achieve chemical
ignition, inflammation, and transition to turbulent jet diffusive combustion, fed
by the continuing discharge of hydrogen. Both experiments and simple transient
shock theory along with chemical kinetic ignition calculations are used to support
interpretation of observations and qualitatively identify controlling gas properties
and geometrical parameters. Although the phenomenon is demonstrated for pressurized hydrogen burst disk failures with different internal flow geometries, similar phenomena apparently do not necessarily occur for sudden boundary failures
of storage vessel or transmission piping into open air that have no downstream
obstruction. However, subsequent reflection of the resulting transient shock from
surrounding surfaces through mixing layers of hydrogen and air may have the
Received 19 July 2005; accepted 3 January 2006.
This work was supported by discretionary research funds available to Prof.
Dryer and by a Mechanical and Aerospace Engineering Department summer
research fellowship (J.N.S.). The authors wish to acknowledge Prof. C.K. Law
for recommending several references cited in the final manuscript revision.
Address correspondence to [email protected]
Downloaded By: [Princeton University] At: 19:20 15 March 2007
664
F. L. Dryer et al.
potential for producing ignition and continuing combustion. Much more experimental and computational work is required to quantitatively determine the envelope of parameter combinations that mitigate or enhance spontaneous ignition
characteristics of compressed hydrogen as a result of sudden release, particularly
if hydrogen is to become a major energy carrier interfaced with consumer use.
Similar considerations for compressed methane, for mixtures of light hydrocarbons and methane (simulating natural gas), and for larger carbon number hydrocarbons show similar autoignition phenomena may occur with highly compressed
methane or natural gas, but are unlikely with higher carbon number cases, unless
the compressed source and=or surrounding air is sufficiently pre-heated above
ambient temperature. Spontaneous ignition of compressed hydrocarbon gases is
also generally less likely, given the much lower turbulent blow-off velocity of
hydrocarbons in comparison to that for hydrogen.
Keywords: Hydrogen; Ignition; Natural gas; Pressure release; Safety; Shock
INTRODUCTION
The transmission, storage, and use of compressed flammable gases are
integral to the industrial, commercial, and residential fabric of the nation.
The compressed flammable gas most familiar to consumers is natural gas.
The use of natural gas continues to grow in refining, power generation,
fertilizer manufacture, and other industrial processing applications as
well as in commercial and residential heating=air-conditioning and food
processing. Compressed natural gas also continues to be of interest as
a transportation fuel alternative.
Currently, the distribution of natural gas resources within North
America is principally by pipelines operating at pressures from 200 to
1500 psi (i.e., 14 to 100 atm, approximately) (NaturalGas.org, 2004).
Local storage is typically on the low pressure side of this range, with
reduced pressures utilized in local commercial and residential distribution
and much lower pressures actually utilized within such facilities. Forecasts suggest that natural gas supply and transport will have to grow substantially to meet a projected need of about 30 trillion cubic feet by the
end of this decade (INGAA Foundation, 2004). The general consensus
is that the traditional basins that today supply about 80% of production
in the United States and Canada are in decline in both absolute and market share terms. There will be an increasing need to replace both the
present producing basins with currently untapped supplies from areas
that are generally more remote from the consuming markets in North
America. Liquefied Natural Gas (LNG) delivered to large marine terminals will contribute heavily to supplies in the next decade. It is likely that
not only will new pipelines be needed to distribute the ever increasing
amounts of natural gas needed, but that the average operating pressure
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
665
of pipelines will increase. For example, long-distance transfer pipelines to
transport Alaskan natural gas are proposed with design pressures in
excess of 140 atm ( 2000 psi) (Transport Canada Prairie and Northern
Region, 2003). The interface with consumers in terms of the use of compressed natural gas as a transportation fuel also typically utilizes storage
pressures of 140 to 200 atm ( 2000 to 3000 psi).
Safety issues involving natural gas at the consumer interface have
been extensively studied and are a subject of continuing interest. For
example, the search continues for improved odorants that are not
removed by soil permeation, and new appliance designs have emerged
that cannot ignite spurious liquid flammable vapors accidentally present.
In all of these consumer interfaces, the layman has little exposure to the
technologies and design criteria underlying their ability to utilize natural
gas safely. While the fundamental issues are essentially transparent to the
end-user, failure to consider the applied implications can have catastrophic consequences.
The flammable gas that is the principal subject of this paper,
hydrogen, also has a long history of importance in the industrial world
in chemical processing, inclusive of refining, fertilizer, and chemical
synthesis of materials. Compared to natural gas, its use in the commercial and residential sectors is very limited. In the last few years, hydrogen has emerged (again) as a favored energy carrier of the future
(Gregory and Pangborn, 1976; Hord, 1978; Fischer, 1986; Kruse
et al., 2002; Shinnar, 2003). Interests are principally driven by the
global climatological implications of carbon emissions from carboncontaining fuels, the promise of increased efficiency of and zero emissions from Proton Exchange Membrane (PEM) fuel cells, and other
issues, including political=societal=strategic hyperbole (Rifkin, 2002).
Converting a significant portion of the present energy infrastructure
to hydrogen would require increasing the production rate up to ten
times the current level, presently on the order of ten million tons annually (Argonne National Laboratory, 2004), and an infrastructure to distribute and locally store hydrogen would need to be developed. While
significant research is underway for finding mechanisms to increase
the energy density of hydrogen storage, liquefaction and high compression will continue to play major roles in hydrogen distribution
and storage. Compressed gas storage at 340–700 atm ( 5000–
10,000 psi) is under consideration by the personal transportation industry. It must be assured that safety at the consumer interface with hydrogen be comparable or exceed that present with natural gas use. Many
government and professional agencies are presently developing design
and product handling standards for the production and transport of
hydrogen that encompass use by consumers in residential and transportation applications.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
666
F. L. Dryer et al.
The potential of unwanted fires and explosions has always been a
major focus of concern with hydrogen. The public sensitivity to hydrogen
relates (perhaps unfairly) to some historical events such as the Hindenburg disaster in Lakehurst, New Jersey on May 6th, 1937 (Vidicom
Media Productions, 2005). The implication of hydrogen in causation of
such accidents derives from its ignition and combustion properties which
are atypical of flammable hydrocarbons such as methane, ethane, and
liquefied petroleum products (e.g., propane, butane). Alcock et al.
(2001) recently summarized some of the properties of both hydrogen
and methane that are important to fire safety (Tables 1 and 2). The most
important points are that the minimum ignition energy of hydrogen is
extremely low, its flammability limits are much wider, its flame temperature is somewhat higher, and its reference laminar flame speed is
far greater in comparison to the respective properties for gaseous
hydrocarbons.
In unconfined, pre-mixed flammable mixtures, slow deflagration can
occur. Furthermore, the slow deflagration can be accelerated (e.g., by
extreme initial turbulence, turbulence from obstacles, or confinement),
and the result will be a transition to a detonation (Berman, 1986; Sidirov
and Dorofeev, 1998; Kaltayev et al., 1999). Once initiated, a detonation is
self-sustaining (i.e., turbulence or confinement are not required) as long
as the combusting mixture is within the detonable range. In comparison
to natural gas, the ability of hydrogen to initiate a detonation or to transition from a deflagration to a detonation (deflagration-detonationtransition, DDT) is more likely.
On the other hand, even though hydrogen has a greater propensity to
leak than other gaseous flammables (Table 2), it has been argued that
should a compressed hydrogen leak occur, the high diffusivity of hydrogen should result in a rapid dispersal of the leak beyond flammability
concentrations. Moreover, the energy content of a hydrogen leak is comparably smaller than that of a methane leak for the same volumetric leak
rate. On the other hand, the turbulent non-premixed flame properties of
hydrogen-air mixtures permit turbulent jet flames to exist at velocities far
in excess of those possible with hydrocarbon-air mixtures (Schefer et al.,
2004).
Finally, the disparity in minimum ignition energy between hydrogen
and gaseous hydrocarbons may not be as significant in terms of hazard
issues as it might appear. The minimum ignition energy for hydrogen=air
mixtures is at near-stoichiometric conditions, i.e., 29% (vol) hydrogen
in air. However, at the lean flammability limit, the ignition energy
requirement is more similar to that of methane, and weak ignition sources
such as electrical equipment sparks, electrostatic sparks or sparks from
striking objects typically involve more energy than is required to ignite
methane and natural gas as well as hydrogen.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
Table 1.
2001)
667
Compressed gas properties for hydrogen and methane (Alcock et al.,
Property
Hydrogen (H2)
Molecular weight
Diffusion coefficient in air at NTP (cm2=s)
Viscosity at NTP (g=cm-s 10 5)
Density at NTP (kg=m3)
Ratio of specific heats, Cp=Cv at NTP
Flammability limits (vol. % in air)
Lower limit (LFL)
Upper limit (UFL)
Minimum ignition energy (mJ)
Adiabatic Flame Temperature in air (K)
% Thermal energy radiated from flame
to surroundings
Maximum burning velocity (m=s)
Concentration at maximum (vol. %)
Burning velocity at stoichiometric (m=s)
Concentration at stoichiometric (vol. %)
Detonability limits (vol. % in air)
Lower limit (LDL)
Upper limit (UDL)
Methane (CH4)
2.016
0.61
89
0.0838
1.308
16.043
0.16
11.7
0.6512
1.383
4
75
0.02
2318
5 to 10
5.3
15
0.29
2158
10 to 33
3.46
42.5
2.37
29.5
0.43
10.2
0.42
9.5
11–18
59
6.3
13.5
Normal temperature and pressure are defined as 20C and 1 standard atm
pressure.
Typical values for different flame configurations and combustion efficiencies.
Table 2.
Relative leak rates of hydrogen and methane (Alcock et al., 2001)
Configuration
Hydrogen (H2)
Methane (CH4)
Relative leak rates (volumetric)
Subsonic flow
Diffusion
Laminar flow
Turbulent flow
Sonic flow
1
1
1
1
0.26
0.77
0.35
0.34
Relative leak rates (energy basis )
Subsonic flow
Diffusion
Laminar flow
Turbulent flow
Sonic flow
1
1
1
1
0.87
2.66
1.18
1.14
Based on lower heating values.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
668
F. L. Dryer et al.
Hydrogen flames are also vastly different from hydrocarbon flames
in terms of heat transfer characteristics (Table 1). With large scale, diffusively limited burning for hydrocarbon fuels (even methane), soot formation can occur, significantly enhancing radiant heat transfer to
surrounding materials. Even without soot formation, hydrocarbon
flames remain somewhat visible because of chemiluminescent emissions.
On the other hand, a hydrogen flame is almost invisible even though
the adiabatic flame temperature is higher. While the thermal radiation
emitted from diffusively limited combustion is considerably lower than
that from a sooting hydrocarbon flame, inhalation of hot combustion
products from hydrogen flames can severely damage lung tissue. Additives to make hydrogen flames luminous and=or that can be used to odorize hydrogen to detect leaks, and yet not impact fuel cell operation, may
be difficult to identify.
As a result of the intense interest of hydrogen as an energy carrier,
there have been numerous studies initiated, some completed, and others
ongoing to refine the understanding of hydrogen safety issues and to
embody this knowledge in design and safety codes (e.g., see National
Hydrogen and Fuel Cell Codes and Standards Coordinating Committee
Website, 2005). It has been said that shut-off safety systems are sufficient
to ensure safety in fuel cell cars and hydrogen fueling stations as they
would avoid unwanted leaks (California Hydrogen Highway, 2004). This
technology is not as amenable to unwanted discharges in pipeline distribution systems or storage vessels. In general, the recent safety evaluations
describe hydrogen as ‘‘a safer material than gasoline or liquefied
petroleum gases’’ (Thomas, 1997; California Hydrogen Highway,
2004), report that general hydrogen handling and safety concerns have
been historically shown to be well in hand (NFPA, 2005), and conclude
that educating the public of the safety issues involved specifically with
hydrogen is very significant in achieving acceptance of hydrogen in
the consumer market (California Hydrogen Highway, 2004; National
Hydrogen Association, 2005).
Acknowledging that unwanted discharges can occur in consumer
handling situations and that sudden fracture of containers or pipelines
can also occur, either in accidents or as subversive acts, a number of
efforts have considered the character of flammable plume envelopes generated by hydrogen leaks (Swain, 2004; Schefer et al., 2004; Houf and
Schefer, 2004). Clearly, the separation of ignition sources from the region
in which flammable mixtures might be produced is generally an effective
safety consideration.
However, the literature contains numerous fire accident scenario
descriptions in relation to hydrogen releases into air that remain, at best,
speculatively understood (Bond, 1991; National Research Council, 1995;
Edeskuty and Stewart, 1996), i.e., there was no clearly identifiable
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
669
ignition source. Frequently, hydrogen’s low minimum ignition energy
and potential catalytic activity are involved in attempting to describe
how these accidents occurred. In the course of preparing the final manuscript, an interesting review on this subject was presented in September
2005 by Astbury and Hawksworth (2005).
The present combustion laboratory had such a fire incident in the
early 1980s when a pressure relief burst disk in a hydrogen gas supply system ruptured as a result of a regulator failure on a 1-A compressed
hydrogen bottle (supply pressure > 40 atm). No readily identifiable
ignition source was apparent in the vicinity of where the fire occurred,
yet the sudden pressure release by the burst disk resulted in a large, turbulent jet diffusion flame, stabilized at the (1=400 stainless steel tube) discharge of the safety system.
Much speculation concerning less obvious ignition sources for
hydrogen leaks fills the technical literature. The Joule-Thompson coefficient for hydrogen becomes negative above a relatively low temperature (28 to 200 K, depending on pressure) and a constant enthalpy
expansion from high pressures under these conditions leads to an
increase in gas temperature (Zuttel, 2003). Joule-Thompson heating is
typically incapable of significantly raising the temperature of hydrogen.
A temperature increase of only about 6 K occurs over a depressurization
from 140 atm (Compressed Gas Association, 2004). Other suggested
ignition sources include static electrical discharge and high velocity particle impact (Edeskuty and Stewart, 1996). The high critical flow velocity
accompanying decompression of high pressure hydrogen can also generate static electrical charge triboelectrically. Impact of small particles carried in such as stream could contain sufficient kinetic energy so as to
produce ignition by their impact, or perhaps catalytic nature. Edeskuty
and Stewart (1996) suggest that total elimination of ignition sources from
static electrical and flowing particles is very difficult under field conditions. But the fact that such ignition sources must also be coincident
with flammable mixtures for ignition to occur seems to be ignored in
many such speculations.
Finally, Edeskuty and Stewart (1996) mention the possibility of
shock wave ignition of premixed hydrogen-air mixtures, without further
discussion of how premixed mixtures and shock wave interactions might
occur. The ‘‘diffusion ignition’’ processes of unmixed flammable gas=air
mixtures with subsequent inflammation and turbulent jet combustion
have escaped much of the applied fire-safety literature, but appear to
have been first studied scientifically by Wolanski and Wojciki (1972).
When a compressed hydrogen source is released into a surrounding oxidizer, there is the potential for development of a transient shock wave
formation. The resulting temperature of the compressed oxidizer can considerably exceed that of the expanding hydrogen, and if mixing of the hot
Downloaded By: [Princeton University] At: 19:20 15 March 2007
670
F. L. Dryer et al.
oxidizer and expanding hydrogen is sufficiently fast in contact regions, it
is possible to obtain, autoignition, flame spreading through the flammable mixture, and transition to turbulent diffusive burning. Wolanski
and Wojciki identified diffusive mixing as the limiting process, and
referred to the ignition scenario itself as ‘‘diffusive ignition.’’
Wolanski and Wojciki (1972) demonstrated diffusive ignition by
studying the issuance of a hydrogen jet into oxygen. A high pressure
hydrogen source was created by using a shock tube apparatus and the
high-pressure hydrogen ruptured a second membrane, resulting in the
ejection of hydrogen through the side wall of a cylindrical test chamber
containing oxygen. A simple diffusion theory was developed to correlate
the dependence of the measured ignition delay as a function of the
resulting shock wave propagating in oxygen. While the inverse problem
was also studied, no demonstration of the phenomenon was performed
for other geometries or with air as the oxidizing media. In comments
to the paper, Li~
nán described a high activation energy asymptotic analysis (later published by Liñán and Crespo, 1976), which considered in
more depth the possible configurations of the evolving mixing layer
separating a flammable gas and oxidizer.
It has recently come to the authors’ attention that in studies of flammable cloud formation by pressurized releases of hydrogen or methane
jets into air, Chaineaux et al. (1990) noted what they termed ‘‘spontaneous ignition’’ (autoignition=inflammation=sustained turbulent diffusive burning) of a compressed hydrogen release into air. In their
experiments, the flammable gas was forced through a small circular hole,
in some cases with a short length of tubing attached downstream. They
reported that spontaneous ignition occurred in only one experimental
configuration repeatedly and only with hydrogen as the test gas. The case
was a 100 atm hydrogen source discharged through a 12 mm hole with
a 120 mm length of 15 mm inside diameter tubing attached downstream
of the hole. ‘‘Nearly instantaneous’’ spontaneous ignition occurred after
the jet was released, with good repeatability for this configuration. No
additional tests of geometrical effects or analyses of the data are reported
to further characterize under what conditions spontaneous ignition might
occur. Through the remainder of this paper, we will use the term ‘‘spontaneous ignition’’ to refer to autoignition and inflammation leading to
sustained turbulent diffusive combustion.
In reviewing the ignition sources possibly involved in the fire accident
that this laboratory incurred, one of the present authors (FLD) theorized
at the time that the spontaneous ignition event might have been initiated
by transient shocks produced by the failure of the burst disk itself. But no
reference to the spontaneous ignition problem or designs to avoid the
problem were found in the hydrogen storage and plumbing standards.
In fact, the most recent hydrogen safety studies and regulations (e.g.,
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
671
NFPA 55, 2005) also fail to discuss the potential for the sudden failure of
a pressure boundary separating compressed hydrogen and air to cause
spontaneous ignition events.
In summary, the data in the literature on spontaneous ignition of
compressed hydrogen releases at the time of the present work was relatively sparse. In fact, no experimental work exists on spontaneous
ignition of pressurized flammable gas releases (of varying geometries)
into an unconfined atmosphere, which would be representative of incidents reported in industry. The present work was undertaken initially
as a senior thesis by two of the authors (Alpert and Homer, 2005) to demonstrate spontaneous ignition of hydrogen-air mixtures solely as a result
of a sudden, high pressure discharge into air. Further experiments and
analyses were subsequently conducted and are reported here, along with
additional analyses of the observations. The experiments clearly confirm
the importance of compressed gas pressure and temperature as well as
geometry on the production of spontaneous ignition in high pressure
hydrogen releases into air. Observations have immediate and significant
implications on the safety measures and piping=storage codes since the
use of high pressures in the transport, storage, and future consumer uses
of compressed hydrogen inherently includes this ignition potential, and
point to the need for additional research to develop appropriate safety
standards.
CONFIRMING EXPERIMENTS
The principal purpose of the present experiments was to confirm that
transient shock formation and mixing associated with failure of a burst
disk separating compressed hydrogen and air can cause spontaneous
ignition of the escaping hydrogen. A variety of experiments were conducted to investigate the dependence of observations on the failure pressure
and internal flow geometry downstream of the disk. Experiments were
performed using different open-ended flow geometries downstream of a
standard commercial, screw union type, straight-through burst disk
holder with 1=200 NPT female connections (1.83 cm ID) on both sides
of the holder (Figure 1a, Figure 1c). Both commercial and in-housemanufactured burst disks with different failure pressures were used in
the experiments. Failure of each burst disk was induced by slowly raising
the flammable gas supply pressure to the burst disk failure pressure.
For safety reasons, a majority of the experiments, particularly those at
failure pressures higher than about 40 atm ( 600 psia) were performed
in open air, out of doors. Ambient air temperatures for the experiments
ranged between 280 K and 305 K, with relative humidity levels between
60 and 90%.
F. L. Dryer et al.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
672
Figure 1. Burst disk holder styles. (a) Union (screw) holder with 1=200 NPT
female fittings; (b) ‘‘T’’ union (screw) holder with 3=800 NPT inlet, 1=400 outlets;
(c) expanded view of setup shown in (a).
As noted below, flow geometry downstream of the burst disk
location was found to strongly influence observations using the
straight-through holder, particularly at lower compressed hydrogen
pressures. Several experiments were also conducted using commercial
burst disks and a ‘‘T’’ type commercial burst disk holder. The ‘‘T’’ type
unit (Figure 1b) had a 3=800 NPT (1.3 cm ID) entrance and a 90 degree
baffle exit with four orthogonally placed exit holes of 6.35 mm diameter
each. No spontaneous ignition events were observed with either compressed hydrogen or natural gas for failure pressures less than
56.6 atm (81 psig). Additional experiments were not conducted with this
burst disk holder design primarily because it was difficult to modify
downstream geometry. No experiments with any type of burst disk
holder were conducted at failure pressures above 113.2 atm (1650 psig).
The present experiments were also formulated specifically to investigate
under what conditions spontaneous ignition might occur. It is probable
that local autoignition might have occurred in some experiments, but did
not transition to turbulent diffusive burning. Characterization of such
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
673
circumstances would require much more sophisticated, time resolved
diagnostics.
A simple apparatus was constructed to study burst disk failure using
the straight through holder (Figure 2). Compressed gas pressure was controlled in a system volume upstream of the burst disk using a two stage
regulator mounted on a 1-A compressed gas cylinder. The compressed
gas cylinder valve and the regulator valve provided redundancy to a supply valve that was used to control compressed gas flow to the system. The
system volume upstream of the disk was varied by incorporating stainless
steel tanks of different volumes from 2.25 liters to 0.25 liters and=or by
using 1=200 NPT stainless steel pipe (1.83 cm ID) of varying lengths (as
little as 0.30 m). Upstream volume had no significant influence on the
spontaneous ignition observations, but varied the overall blow-down
time subsequent to disk rupture. Essentially the spontaneous ignition
process (when it was observed) occurred on very short time scales in comparison with system depressurization times. As noted above, flow
geometry downstream of the burst disk holder was varied in the experiments and several tests were repeated for specific configurations.
Threaded pipe fittings were assembled with Teflon pipe ribbon, and all
fittings were leak checked with soap solution for each configuration prior
to each experiment.
Commercial ‘‘Type B’’ aluminum burst disks (Figure 3) manufactured by BS&B Safety Systems, LLC were used in initial experimentation.
The Type B design, developed more than sixty years ago, is a solid metal
disk that retains its initial contour during exposure to normal system
Figure 2.
Schematic of experimental configuration.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
674
F. L. Dryer et al.
pressure differences. The outer collar is form-fit around the union and
screw flanges and the inner section of the disk forms a concave ‘‘crown’’
extending downstream in the holder (Figure 1c). The disk material acts as
a metal gasket seal when the union flanges are tightened. Thin layers of
silicone grease were also applied to the outer edges of the disks to further
aid in achieving the seal with the union flanges. An upstream overpressure build-up above the rating of the disk causes a thinning of the
metal at the center of the concave crown. Failure then takes place near
the center, and the subsequent release of the compressed gas through
the holder results in an opening pattern of varying shapes, with ‘‘petals’’
of disk material expanded to the outer diameter of the holder passage.
Under most instances, the petals are entirely retained on the remaining
disk (Figure 3).
Commercial burst disks are reasonably expensive, and thus Type B
burst disks were also manufactured in the laboratory using different
thicknesses of aluminum shim stock from 76–229 mm (0.003–0.00900 )
and a two-step process. Pieces of shim stock were cut to size. Silicone
spray was applied to the sized flat shim stock disks, and the disks were
then placed in a spare screw union disk holder. The screw union
was wrench tightened to form the flat shim stock circumference into
the geometrical shape of the commercial disk crown. An aluminum mold
was numerically machined to the concave shape of the center of the
commercial disk crown, and in a second process step, the shaped shim
stock disk was pressed in the mold to obtain the same overall manufactured disk shape as that found for a commercial Type B disk.
Figure 3.
A ruptured Type B burst disk.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
675
Large changes in the average failure pressure of manufactured disks
were obtained by using different shim stock thicknesses. For the thinnest shim stock, small adjustments of failure pressure (< 1.5 atm) were
made by modifying the second step and producing small variations in
the concave deformation in the center of the crown. The failure pressure from disk to disk for each shim stock and manufacturing method
was nearly as repeatable ( 0.3 atm) as the designed failure pressure
of commercially manufactured disks. No differences in observations
were noted between results obtained with commercial or manufactured
disks.
Overall, more than 200 experiments were conducted with hydrogen,
the straight-through screw union burst disk holder and various upstream
and downstream geometries for failure pressures ranging from 11.2 atm
to 113.25 atm. In each experiment, compressed gas was purged through
the entire system upstream of the disk at low pressures ( 1.5 atm), with
the burst disk holder screw union fittings only hand-tightened. During the
purge event, the burst disk holder screw union fittings were wrench-tightened, and then the purge exhaust valve was closed, leaving the system
pressurized. Leak checks of the system were conducted using soap solution and by monitoring the system pressure with the supply valve closed.
The supply valve was reopened and the compressed gas system pressure
was then slowly raised by increasing the two stage regulator pressure setting manually until the burst disk failed. Alternatively, the regulator
pressure was set above the failure pressure, the supply valve was opened,
and the rate of system pressure increase was controlled by a needle valve
placed in the line just downstream of the supply valve.
In all cases, the system pressure was monitored using calibrated
pressure test gages with appropriate pressure scales (< 0.2 atm accuracy). Test gage readings at the time of disk rupture were monitored visually in early experiments and, in later experiments, recorded digitally
either using a calibrated pressure transducer (P < 69 atm, 1000 psig) or
a video record of the test gage reading and event. In tests where the pressure was recorded using a pressure transducer, compressed gas temperatures in the system storage volume were also recorded using a small
diameter, sheathed thermocouple. Pressurization rates were sufficiently
slow that initial compressed gas temperatures never varied from ambient
air temperatures by more than 10 K.
The maximum hydrogen flow rate from the regulator to the system
volume upstream of the burst disk after burst disk failure was restricted
in some of the experiments using a needle valve in the supply line, and in
all of the experiments by the 1=400 stainless steel tube flow resistance. The
supply valve was closed several seconds after the burst disk failed. The test
sequence was then repeated with a new burst disk and the same or a
modified downstream flow configuration.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
676
F. L. Dryer et al.
In early experiments, it was quickly discovered that the flow
geometry downstream of the burst disk holder had strong effects on
whether and at what minimum burst disk failure pressure spontaneous
ignition would occur. These initial experiments were conducted with
the downstream flow geometry shown in Figure 4, threaded into the exit
of the straight through burst disk holder. The minimum internal exit
diameter of the male Swagelok fitting is 0.472 cm (0.18600 ), and the overall
length of the region from the plane of the burst disk to plane where area
reduction occurs was 5.08 cm (200 ).
Figure 5a documents a spontaneous ignition event and the ensuing
turbulent jet flame established by the failure of a commercial burst disk
at 43.04 atm pressure (618 psig). Though hydrogen combustion itself is
almost non-luminous, small levels of impurities in the surrounding air
and within the flow system cause the minimal luminosity observed in
Figure 5a. Noise associated with burst disk failure and spontaneous
ignition of the gases inside the downstream flow cavity was always substantially louder than the noise associated with a disk failure and no
spontaneous ignition. The significantly louder noise is indicative of the
spontaneous ignition process occurring inside the downstream flow configuration and likely results from deflagration=detonation transition in
the flammable mixtures within the cavity. The ensuing stabilized turbulent jet diffusion flame downstream of the fittings also resulted in a distinct roaring sound not present with free-flowing, unignited hydrogen.
Tests were repeated with the same configuration and commercial burst
disks at failure pressures of 56.6, 42.5, and 22.1 atm (817, 610, and
Figure 4. Downstream flow geometry used in initial experiments, composed of a
1=200 NPT pipe nipple 5.08 cm (200 ) long, a 1=200 to 3=800 NPT female brass reducing union, and 3=800 NPT male to 1=400 male Swagelok tube reducer.
677
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
Figure 5. Sustained turbulent diffusion flame of hydrogen and air ignited by failure of a burst disk at (a) 43.1 atm (618 psig) using the downstream geometry
shown in Figure 4 and (b) 86.1 atm (1250 psig) with a short length of unrestricted,
constant diameter pipe (12.7 cm, 500 ).
310 psig), with identical results. No spontaneous ignition was observed
with commercial burst disks designed to fail at 14.4 atm (197 psig).
On the expectation that a minimum failure pressure should exist
below which no spontaneous ignition would occur, the test protocol
using the configurations in Figures 4 and 5a was repeated with 15 different in-house manufactured burst disks designed with average failure
pressures between 15.6 atm and 26.9 atm (215 to 380 psig). The failure
pressure in each test was obtained from video records of the experiments, including the test pressure gage readings. Spontaneous ignition
events were statistically distributed with a zero probability for failure
pressures less than 20.4 atm (285 psig), and certain inflammation for
Downloaded By: [Princeton University] At: 19:20 15 March 2007
678
F. L. Dryer et al.
failure pressures above 22.1 atm (310 psig). In addition to identifying a
minimum pressure range below which no spontaneous ignition was
observed, it was found that reducing the length of the 1=200 NPT pipe
nipple to 3.81 cm (1.500 ) led to no spontaneous ignition at all failure
pressures even up to 56.6 atm.
Modification of the flow geometry downstream of the 1=400 reducer
tube exit also affected observations. For example, the addition of a
straight section of 1=400 stainless steel tubing (6.35 mm OD, 4 mm ID)
downstream of the reducer fitting up to 46 cm long continued to result
in inflammation for the same failure pressures without tubing. Addition
of tube lengths greater than 300 cm suppressed inflammation altogether
at failure pressures as high as 55.4 atm (800 psig). Apparently sufficient
heat can be removed from gases ignited internally by the shock discharge
leading to quenching before the inflamed volume can exit the tube.
In further experiments with reducer fittings that had internal exit diameters larger than that in Figure 5a, we found that the minimum pressure
above which spontaneous ignition was observed increased. For example,
for configurations with a reducer fitting (1=200 NPT male to 1=200 male
tube) with an exit of 1.305 cm ID (0.47500 ID), a minimum failure pressure
of more than 67.3 atm (975 psig) was required to consistently produce
spontaneous ignition. From such experiments and the simple theory discussed next, it was hypothesized that above some higher minimum disk
failure pressure, spontaneous ignition might occur without any downstream fittings. Experiments were then conducted without any downstream geometry using the standard burst disk union flange (Figure la,
Figure lc). No spontaneous ignition events were observed for in-house
manufactured disk failure pressures up to 113.2 atm (1650 psig). It was
theorized that the transient mixing of expanding hydrogen with shocked
hot air required to achieve spontaneous ignition is clearly related to mixing issues caused by transient flow conditions in the downstream pipe sections. As a result, a large number of experiments were performed using
1=200 NPT pipe (1.27 cm ID) of different lengths (3.81, 4.54, 12.7,
15.24, 25.4, 30.48 cm) and different disk failure pressures. In some of
the tests the pipes were mated using 1=200 NPT female pipe unions
(1.83 cm ID) to obtain longer lengths.
The overarching conclusion of these experiments was that achieving
spontaneous ignition requires downstream geometry and failure pressure
combinations that provide a critical level of sensible enthalpy in the shock
heated air to cause autoignition of mixtures formed with hydrogen, as
well as a critical volume of mixed material that is capable of undergoing
flame spreading. For example, Figure 5b shows a spontaneous ignition
event observed with a disk failure pressure of 86.1 atm (1250 psig) for a
constant cross-section 1=200 NPT exit pipe, 10.16 cm (400 ) long. The
overall cavity length from burst disk plane to the installed pipe exit
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
679
was 12.7 cm (500 ). For this disk failure pressure, spontaneous ignition was
observed for downstream lengths up to 90 cm, composed of 1=200 NPT
stainless steel pipe section, each 30.5 cm (1200 ) in length, mated with
1=200 NPT female unions. No spontaneous ignition occurred in experiments with single pieces of pipe (i.e., no mating unions) with overall cavity lengths from burst disk plane to exit of less than 10.16 cm (400 ).
However, if shorter lengths of pipe were connected together with pipe
unions to produce an overall length less than 10.16 cm, spontaneous
ignition resulted. Apparently, the internal diameter area change resulting
from the presence of the 1=200 NPT union introduced the additional
mixing necessary to achieve spontaneous ignition. These observations
suggest that interactions of the core transient flow with the transient,
developing boundary layer within the downstream geometry contribute
significantly to achieving sufficient molecular mixing of expanding
hydrogen and shocked air to cause spontaneous ignition. Even though
the temperatures of mixed hydrogen and air may be high enough to
achieve local ignition, a critical volume of flammable mixture must also
exist in order to establish flame spreading and spontaneous ignition.
Similar results were obtained for disk failure pressures of 99.6 atm and
113.2 atm (1450 and 1650 psig). The minimum constant area pipe length
required to achieve ignition only slightly increased with increasing disk
failure pressure, and did not decrease, as would be expected if the ignition
phenomena was limited by chemical kinetic ignition delay time rather
than mixing.
Finally, compressed methane release was tested in a cursory fashion
under a number of configurations where experiments with hydrogen
resulted in spontaneous ignition In these tests, 99.98% pure compressed
methane was used with the knowledge that the ignition delay for methane=
air mixtures are typically equal to or longer than those for natural gas
(methane that typically contains small amounts of higher hydrocarbons
such as ethane, propane, and butane). No spontaneous ignition cases
resulted for any configuration, with and without reducer exit fittings for
failure pressures as high as 99.6 atm (1450 psig).
In summary, the experiments show that compressed hydrogen suddenly released into air by failure of a burst disk can produce spontaneous
ignition if the burst disk failure pressure is sufficiently high, and the flow
geometry after the burst disk results in sufficiently fast mixing of the
escaping hydrogen and shocked air in contact with the expanding hydrogen. The range of failure pressures and geometries corresponding to
spontaneous ignition events identify that reflected shock, shock-shock,
and perhaps even shock focusing can be key in achieving sufficiently
short chemical ignition delay times, while downstream geometry also
contributes in enhanced mixing rates of shocked air and expanding
hydrogen that produce sufficient volumes of flammable mixture in which
Downloaded By: [Princeton University] At: 19:20 15 March 2007
680
F. L. Dryer et al.
an ignition kernel can result and propagate to establish a flame. At low
compressed gas pressures, the chemical ignition time becomes the limiting
factor, while at high compressed gas pressures, the mixing time of
shocked air and expanding hydrogen to achieve significant volumes of
flammable mixture becomes the limiting factor.
Downstream geometries that produce sufficient mixing and generate
shock reflections can induce spontaneous ignition at lower compressed
gas pressures, while at higher compressed gas pressures, transition of
ignition to inflammation is primarily dependent on forming a critical
ignition kernel that can spread a flame through the remaining flammable
mixture. Boundary layer interactions of the expanding hydrogen inside
the downstream pipe geometry appear to be important in achieving sufficient mixing of shock heated air and hydrogen in order to produce
spontaneous ignition. Finally, by changes in downstream geometry alone,
spontaneous ignition can be avoided altogether, even for high-pressure
releases that clearly produce shock temperatures sufficient for autoignition to occur. The present observations cannot preclude that local autoignition might have occurred, but could not transition to turbulent
diffusive burning.
The present experiments do not encompass the range of pressures
under discussion for hydrogen storage (up to 681 atm, i.e., 10,000 psi).
Much more detailed experimental studies are needed to quantitatively
determine the relative importance of different mixing mechanisms,
shock-shock, and shock-focusing phenomena on producing spontaneous ignition. Future work should include much higher pressures,
different failure diameters and down stream geometries, as well as
consideration of boundary failures into open air along with shock
interactions with surrounding obstructions in close proximity to the
boundary failure.
Next, some simple theoretical arguments are offered to further support the experimentally observed spontaneous ignition events which are
consistent with transient shock induced autoignition limited by local mixing phenomena between flammable and non-flammable materials, and to
further extend to the observations reported with hydrogen to small molecule hydrocarbon discharges into air. Certainly, the observed spontaneous ignition dependencies on pressure and geometry as well as
their reproducibility are inconsistent with electrostatic discharge or particle impact phenomena as ignition sources, frequently used to explain
ignition of hydrogen releases into air. Some accidents and observed spontaneous ignition events may indeed be related to these mechanisms, but
they are not the controlling phenomena in the present experiments. For
reference, sample video records of some of the hydrogen spontaneous
ignition tests described here are available on the web (Fuels and Combustion Research Laboratory, 2005).
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
681
SIMPLE TRANSIENT SHOCK THEORY AND IMPLICATIONS
Hydrogen-oxygen combustion and explosion limits are a classic problem
(Lewis and von Elbe, 1987). Historically, transient shock ignition
phenomena have been utilized to generate conditions to study the
ignition and combustion of homogenous premixed mixtures of gaseous
fuels and oxidizers. Extensive literature exists on single pulse and
reflected shock tube studies of premixed hydrogen-oxygen, methaneoxygen, and natural gas-oxygen mixtures (e.g., Eubank et al., 1981;
Blumenthal et al., 1996; Huang et al., 2004). As noted earlier, much less
work has addressed transient shock phenomena produced by compressed
gas releases into air as a potential spontaneous ignition source. In this
case, both the shock strength to produce sufficiently short ignition delays
in mixed gases as well as sufficient mixing of shocked air and expanding
flammable gas are required to achieve spontaneous ignition The mixing
processes involved in the events observed above are limited by macroscopic mixing which enhances the interface between shock heated air
and expanding hydrogen (well beyond that associated with the initial
burst disk rupture) such that sufficient molecular mixing can occur.
While more complex numerical simulations of transient shock processes can be applied to this problem (e.g., Wang and Rehm, 2001; Liu
et al., 2006), the pressures and temperatures produced by transient gas
dynamic effects can be estimated using ideal shock equations (Ames
Aeronautical Laboratory, 1953; Shapiro, 1954) to qualitatively estimate
limiting chemical ignition times as a function of flammable gas and air
properties (initial temperatures, pressures, flammable gas molecular
weight).
Qualitatively, the conditions are shown at time t1 in Figures 6 and 7.
The pressure and temperature in region 2 behind the transient shock are
strong functions of the speed of sound (i.e., molecular weight, temperature) of the compressed (flammable) driver gas relative to that of the driven gas (air), as well as the driver gas initial pressure in region 4, P4, in
comparison to the pressure in region 1, P1. By proper selection of initial
parameters and gas properties, temperatures and pressures created in
region 2 can be sufficiently high to induce chemical reactions in mixed
regions between the driver and driven gases (contact surface). Chemical
reaction in the contact region requires both sufficient mixture temperatures to induce ignition as well as molecularly mixed regions containing
both hydrogen and air. That flame propagation away from the ignition
region should occur will require that these mixed regions lie within
flammability limits and are of some critical volume that provides
sufficient energy release to sustain flame propagation.
Setting aside the critical volume issue for the moment (assuming
that mixing occurs sufficiently fast to result in mixtures that lie within
F. L. Dryer et al.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
682
Figure 6. Schematic of classical single pulse shock tube operation. (a) Initial
conditions; (b) time t1, after fracture of diaphragm; (c) position of contact surface,
shock as a function of time. (From Shapiro, 1954.)
flammability limits), 1-D shock analyses can be utilized to estimate spatial time scales based upon contact surface, shock velocities and cavity
length from the burst disk to the exit plane used in the experiments.
The chemical ignition time will be a function of the flammable gas and
air concentrations in the mixture regions. The mixture region temperatures will be a function of the flammable gas properties, pressures P4,
P1 and temperatures T4, T1. The local fluid temperatures in mixed regions
can be no greater than the predicted T2 and must be less as a result
of mixing of the expanding (cooled) hydrogen with the shock-heated
air. Qualitative estimates of mixture temperature can be generated
by assuming that mixing (by whatever means) is instantaneous and
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
683
Figure 7. Pressure and temperature as a function of spatial location at time t1
(Figure 7).
adiabatic. Assuming equal molar specific heats for hydrogen and air, the
lean flammability limit mixture would be on the order of 30 to 50 K
below T2, and stoichiometric mixtures would be 240–450 K below T2.
Rich flammability limit mixtures would be the order of 600 K below T2.
Using hydrogen as the driver gas and air as the driven gas, T2 can be
estimated as a function of P4 from simple 1-D shock theory, yielding the
results shown in Figure 8b. Using an available kinetic solver (Lutz et al.,
1987), and a recent detailed chemical kinetics and thermochemistry
model for hydrogen and oxygen developed in this laboratory (Li et al.,
2004), the ignition delay of premixed hydrogen-air mixtures at the lean
and rich flammability and stoichiometric mixture conditions were calculated for the predicted P2 and T2 (Figure 8a). The calculations shown in
Figure 8a were performed using a constant volume assumption with the
chemical ignition delay time determined as the time to pressure rise. This
is a common approach for calculating shock tube ignition delay times for
premixed gas mixtures (Blumenthal et al., 1996). A constant pressure
assumption and other criteria for defining ignition delay yield times of
very similar magnitude. Also shown in the figure are the time scales for
movement of the contact surface from the diaphragm rupture location
by 1 cm. The calculated ignition time corresponds more closely with
ignition of mixtures near the lean flammability limit since the temperature
for these mixtures differ only slightly from T2, as pointed out previously.
F. L. Dryer et al.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
684
Figure 8. (a) Calculated hydrogen-air ignition delay times (sign) for flammable
mixtures at the temperature and pressure conditions behind an incident normal
shock (T2 and P2) associated with the driver pressure, P4. The time scale for
movement of the contact surface (CS) over a distance of 1 cm is also shown;
(b) temperature behind incident and reflected shock waves (T2, T5) as a function
of driver gas pressure, P4; driver gas, hydrogen; driven gas, atmospheric air; (c)
ignition delays for flammable hydrogen-air mixtures at the reflected shock conditions of T5, P5 associated with the driver pressure, P4 in (b); (d) Ignition delays
for flammable methane-air mixtures at reflected shock conditions (T5, P5).
One can immediately see from Figure 8a that P2 and T2 values associated with a driver pressure of P4 ¼ 22.1 atm (310 psig) cannot possibly
explain the spontaneous ignition observed in the initial experiments
with the smallest exit diameter and pipe length used (see Figure 4 and
Figure 5a). Returning to the 1-D description of the problem, ignition
times even at T2 and P2 are much longer than available time scales
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
685
associated with the 1-D movement of the contact surface over the distance from where the burst disk is initially located to the reducer fitting,
where the area change occurs in the tube internal diameter (Figure 4).
In the experiments, however, the reduction in tube cross-sectional
area results in partial reflection from as well as transmission of the
initial transient shock through the cross sectional area constriction.
The result will be that local temperatures and pressures will be
further increased ahead of the progressing initial contact surface by the
partial reflection of the initial transient shock. As the constriction
cross-sectional area approaches zero, the reflected shock effects will
approach those commensurate with reflection of the transient shock from
a tube end-wall, corresponding to the maximum pressure and temperatures that can be produced in the air ahead of the contact surface. The
maximum reflected pressure and temperature (P5, T5) can again be
estimated using 1-D shock theory for a closed boundary configuration
(Shapiro, 1954).
Figure 8b compares the calculated T2 and T5 as functions of the
hydrogen driver pressure, P4. Because both pressure and temperature
in the mixed regions can be increased as a result of shock reflection,
ignition delay times will decrease significantly, as shown in Figure 8c,
relative to those estimated based upon an initial shock transmission without reflection (Figure 8a). Ignition delays are reduced not only by the
increased temperature, but also by the increased pressure due to shock
reflection. The pressure and temperature conditions (P5, T5) for a
reflected normal shock corresponding to an initial driver pressure of
P4 ¼ 22.1 atm (310 psig) result in a chemical ignition delay estimate at
least two to three orders of magnitude shorter than contact surface transit
time (Figure 8c). This result can only mean that when spontaneous
ignition was suppressed solely by shortening the distance from burst disk
to the cross-sectional constriction, the limiting issue must have been that
a critical volume of mixed flammable gases was not achieved in the physical time available for gas flow development in the region.
The actual pressures and temperatures generated by partial reflection
decrease from those calculated with a closed-end boundary assumption
(P5, T5) as the reducer cross-sectional exit area is increased, eventually
approaching the values produced by the initial transient shock (P2, T2)
for a constant diameter duct. As noted in the experiments, when reducers
are present, the minimum pressure ratio for which spontaneous ignition
will occur should increase as the ratio of the reducer cross-sectional exit
area to the pipe cross sectional area is increased. At pressure ratios above
about 85, no reducer appears necessary to achieve mixture static temperatures and pressures sufficient to cause spontaneous ignition, provided a
sufficient volume of flammable mixture is achieved in the hydrogen transit time within the downstream pipe length.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
686
F. L. Dryer et al.
Of course, the real processes that occur when the burst disk ruptures
are multidimensional. The shock structures produced by the rupture are
affected by both the initial geometry of the burst disk at failure (multidimensional) and the (stochastic) fracture geometry, producing shocksidewall and shock-shock interactions (Jiang et al., 1997). It is possible
through such shock interactions and ‘‘shock focusing’’ (Fay and Lekawa,
1956; Gelfand et al., 2000; Bartenev et al., 2001) exist and produce higher
local temperatures and pressures (and hence shorter ignition delays) than
those estimated on the basis of simple 1-D theory. Moreover shock interactions with the side walls of the cavity will result in heating the air in the
developing boundary layers near the walls and in increasing contact surfaces between the expanding hydrogen core and shock-heated air.
Returning now to the issue of mixing, in order for transient
shock phenomena to produce the spontaneous ignition observed in the
experiments, mixed regions of hot driven gas and expanding, cold hydrogen gas must reach temperatures sufficient to result in autoignition. Subsequent combustion was observed to transition from this partially
premixed state to an external turbulent jet diffusion flame (Figure 5).
What are the mechanisms that contribute to achieving sufficient mixing
on the time scales of these events to result in the experimental observations? For all of the experiments conducted, the characteristic times for
mixing are less than the order of 10 4 s. Molecular diffusion across contact surfaces is orders of magnitude too slow to account for any significant formation of flammable mixture volumes, unless the overall
interface area between expanding hydrogen and shocked air are increased
by the developing flow conditions. Time scales associated with fluid
dynamic instabilities at a 1-D contact surface (which could promote mixing) appear to be too long to be of relevance (Taylor, 1950; Brouillete,
2002) in terms of generating macroscopic mixing. However, recent studies
of membrane failure show that the rupture itself of a planar diaphragm
can produce significant turbulence and mixing at the subsequent contact
surface (Erez et al., 2000). More importantly, multi-dimensional shock
and contact surfaces can interact with the cavity sidewalls downstream
of the burst disk failure, leading to turbulent mixing of shocked driven
gas in the developing boundary layers at the pipe wall with driver gas
in the core flow. Additionally, any change in pipe diameter (union fittings, reducer fittings) can lead to enhanced mixing. Which of these
mechanisms are most important depends on the experimental configuration, but it is clear that interactions with the pipe boundary are important to observations. For example, spontaneous ignition occurred
experimentally when using 1=200 NPT pipe lengths joined by NPT unions,
however, for the same burst pressure, no spontaneous ignition resulted
when using a straight 1=200 NPT pipe with the same length as the joined
pipes.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
687
Figure 9 schematically describes the initial shock formation, turbulent mixing at the contact surface formed as a result of disk fracture,
and the initial shock reflection at the cylindrical side walls downstream
of the disk rupture location, resulting in reflected shock heating of the
fuel=air mixtures formed in the contact region and at the cavity side
walls. As noted above, reflective shock focusing effects may also be
present near the axis of the flow, establishing even higher local temperatures and pressures. Even without considering shock focusing effects,
however, the reflected shock conditions at hydrogen pressures above
22 atm are likely to result in chemical ignition delay times short enough
to induce spontaneous ignition. Limited mixing of shocked air and
hydrogen may still prohibit spontaneous ignition. Based on experimental
results, above hydrogen pressures of approximately 86 atm and for downstream geometries without area changes, spontaneous ignition takes place
as long as sufficient mixing is induced by downstream geometry.
Figure 9. Schematic of the burst disk rupture process and mixing phenomena,
(a) configuration prior to rupture; (b) initial multi-dimensional shock formation;
(c) reflection from straight-through holder downstream cylindrical side wall.
Interaction of reflected shock with contact surface enhances turbulent mixing
and contributes to its further instability; (d) convergence of reflected shocks at
the flow centerline increase pressure and temperature. The light grey shaded
regions denote the growth of a boundary layer of hot air as the shock waves
develop. The flammable gas=air mixtures generated at the boundary layer interface as well as at the contact surface lead to the experimentally observed ignition
events.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
688
F. L. Dryer et al.
In work that was recently presented during the revision of this manuscript, Liu et al. (2006) describe a numerical simulation of hydrogen jetting from a high pressure tank. Direct numerical simulations in this work
showed that for 40 and 70 MPa compressed hydrogen expanding through
a 1 mm hole, chemical ignition occurred at the interface between the
expanding hydrogen and the air heated by the expanding hemispherical
shock. However, no inflammation was observed because sufficient heat
release was not available to overcome the cooling effects resulting from
flow divergence. Essentially the numerically estimated mixing time scales
were too long to produce a flammable mixture of critical volume. The
results of this work suggest that compressed gas releases of hydrogen
from unconfined planar surfaces are unlikely to produce spontaneous
ignition, contrary to the observations reported for the present experiments that involve geometric constraints downstream of the point of
pressure boundary failure. It is interesting to also note that the present
observations are similar to the spontaneous ignition results obtained in
the experiments of Chaineaux et al. (1990) described earlier. The present
experiments not only confirm these results but show that imposition of a
geometric constraint downstream of the pressure boundary failure
location is critical to achieving spontaneous ignition. It is reasonable to
speculate that other external geometries placed in the path of expanding
spherical shocks might also lead to spontaneous ignition for planar pressure boundary failure conditions.
SPONTANEOUS IGNITION OF NON-PREMIXED
HYDROCARBON FUEL/AIR MIXTURES
The strength of the initial transient shock generated by sudden boundary
failure is a function of both P4=P1 and the molecular weight ratio of the
driver and driven gases. Figure 8d presents ignition delay times, based
upon 1-D theory, as a function of reflected pressures and temperatures
(P5, T5) associated with the driver pressure, P4, in the case of methane
as the driver gas (large molecular weight). Note that the change in molecular weight of the driver gas substantially affects the calculated
temperatures achievable for the same pressure. Using 1-D theory and
methane as the driver gas, ignition delay times were calculated using
the GRIMECH-3 mechanism for methane oxidation kinetics (Smith
et al., 1999) as a function of compressed gas pressure. According to
Figure 8d, spontaneous ignition should not be expected, and experiments
confirm that even shock reflection and shock-shock interactions under
the present configuration do not produce spontaneous ignition.
For experiments with pressures as high as 99.6 atm (1450 psig) and
using the configuration of Figure 4, no clear visual sign of spontaneous
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
689
ignition was observed. Moreover, turbulent free-jet diffusion flames
could not be stabilized at steady flow exit conditions at the reducer fitting, even with piloted ignition. Considerably more experimental work
will be necessary to establish if spontaneous ignition of natural gas mixtures with air can be induced by transient shock phenomena associated
with pressure-boundary failure. This work is certainly warranted, given
that storage and transmission pressures are likely to continue to increase.
While simple theory and the potential for shock-shock interactions and
shock focusing all suggest that this phenomenon has a finite probability
for methane and natural gas, transition of ignition to flaming combustion
of the continuing fuel jet will depend on the geometry envelope surrounding site of pressure release and bluff-body stabilization.
Using the simple 1-D shock calculation methods described here,
Figure 10 presents pressures and temperatures for the initial transient
shock (P2, T2) and reflected shock (P5, T5) for hydrogen- and methaneair systems compared with those using ethane or propane as the driver
gas. The change in molecular weight of the driver gas from that of hydrogen to even that for methane has a very significant effect on the supply
pressure, P4, required to achieve values of P5 and T5 that lead to short
chemical ignition delays comparable to time scales associated with shock
and contact surface velocities and the length scales of downstream pipes
and geometries investigated here. While it appears that ethane might be
marginally capable of compression to values of P4 sufficient to cause
ignition and inflammation from shock reflections and focusing, liquefied
gaseous fuels such as propane are not able to achieve the necessary gas
phase driver pressures.
Figure 10. Calculated pressure and temperature as functions of driver gas species and pressure, P4, (a) behind an incident shock (P2, T2); (b) behind a reflected
shock (P5, T5).
Downloaded By: [Princeton University] At: 19:20 15 March 2007
690
F. L. Dryer et al.
CONCLUSION
The present work conclusively establishes that within the storage and
pipeline pressures used today and=or contemplated in the future for
hydrogen, transient shock processes associated with rapid pressure
boundary failure have the capacity to produce spontaneous ignition of
the compressed flammable released into air, provided sufficient mixing
is also present. Pressure boundary failure geometry, multi-dimensional
shock-boundary, and shock-shock interactions in addition to molecular
diffusion can provide the necessary short mixing time scales. Turbulent
free jet hydrogen flames can be stabilized at sufficiently high jet velocities
to result in continued combustion. Experimental observations identify
that the minimum compressed gas pressure (relative to atmospheric
pressure air) at which spontaneous ignition occurs is dictated by reflected
shock and shock-shock interactions. The repeatable nature of the ignition
and its characteristic time scale suggest that proposed ignition alternatives (e.g., electrostatic discharge) are not contributory to the present
observations. Further experiments are needed to quantitatively define
the experimental parameter ranges of gas temperature, pressure, pressure
boundary thickness and structure, and pressure boundary local geometry
that contribute to determining the envelope of design parameters over
which shock induced inflammation will or will not result. These works
need to include the analysis of external boundary failure and shock reflection from surrounding structures to encompass compressed pressure vessel and high pressure pipe failure modes inside enclosures. There is no
doubt, however, that this phenomenon has been a cause of past fire incidents involving sudden releases of compressed hydrogen gas into air (e.g.,
National Research Council, 1995).
Simple 1-D shock calculations suggest that sudden compressed methane and natural gas discharges into air may cause similar reflected shock
and focused shock-shock induced spontaneous ignition, but at considerably higher compression pressures (relative to hydrogen). It is unlikely
that stabilized turbulent combustion of a gas jet will result for compressed hydrocarbon releases unless physical flame stabilization mechanisms (e.g., bluff-body stabilization) are present. The molecular weight
(relative to air components) and=or the liquefaction pressure of the fuel
preclude transient shock induced spontaneous ignition of other compressed hydrocarbon gases. To date, it appears that this phenomenon
has escaped consideration in the analysis of most accidental fires and
in the development of piping and storage safety codes both for hydrogen
and natural gas.
Moreover, an important observation is that downstream flow
geometry or objects in the path of exiting jets formed by the sudden discharge of compressed flammable gases into air have impact on whether
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
691
shock induced spontaneous ignition. Methodologies to enhance or mitigate continued combustion likely exist and such options need to be identified and characterized by more detailed measurements so that safe
storage and distribution systems can be interfaced with general consumer
use of compressed hydrogen.
REFERENCES
Alcock, J.L., Shirvill, L.C., and Cracknell, R.F. (2001) Compilation of Existing
Safety Data on Hydrogen and Comparative Fuels. Report EIHP 2, Fifth
Framework Programme (1998–2002), Shell Global Solutions.
Alpert, J.Y. and Homer, C.J. (2005) Shock Induced Auto-Ignition of Previously
Unmixed Hydrogen=Air Mixtures. Undergraduate Senior Thesis, Mechanical
and Aerospace Engineering, Princeton University, Princeton, NJ.
Ames Aeronautical Laboratory (1953) Equations, Tables, and Charts for Compressible Flow. Report 1135, Ames Aeronautical Laboratory, Moffett Field, CA.
Argonne National Laboratory (2004) Basic Research Needs for the Hydrogen
Economy. Report on the Basic Energy Sciences Workshop on hydrogen production, storage, and use, U.S. Department of Energy, Washington, DC.
[Online] Available: http://www.sc.doe.gov/bes/reports/files/NHE_rpt.pdf
[1 July 2005].
Astbury, G.R. and Hawksworth, S.J. (2005) Spontaneous ignition of hydrogen
leaks: a review of postulated mechanisms. International Conference on
Hydrogen Safety, Pisa, Italy.
Bartenev, A.M., Komik, S.V., Gelfand, B.E., Gronig, H., and Oliver, H. (2001)
Effect of reflection type on detonation initiation at shock-wave focusing.
Shock Waves, 10, 205–215.
Berman, M. (1986) A critical review of recent large-scale experiments on hydrogen=air detonations. Nucl. Sci. Eng., 93, 321–347.
Blumenthal, R., Fieweger, K., Komp, K.H., and Adomeit, G. (1996) Gas
dynamic features of self ignition of non diluted fuel=air mixtures at high
pressure. Combust. Sci. Technol., 113, 137–166.
Bond, J. (1991) Sources of Ignition: Flammability Characteristics of Chemicals
and Products, Butterworths-Heinemann Ltd., Oxford, p. 52.
Brouillette, M. (2002) The Richtmyer–Meshkov instability. Annu. Rev. Fluid
Mech, 34, 445–468.
California Hydrogen Highway (2004) Hydrogen Use and Safety [Online].
Available:
http://www.hydrogenhighway.ca.gov/hysafety/hysafety.htm
[18 May 2005].
Chaineaux, G., Mavrothalassitis, G., and Pineau, J. (1990) Modelization and
validation tests of the discharge in air of a vessel pressurized by a flammable
gas. In Kuhl, A.L., et al. (Eds.) Dynamics of Detonations and Explosions:
Explosion Phenomena, Prog. Astronaut. Aeronaut., Vol. 134, pp. 104–137
Compressed Gas Association (2004) Hydrogen Vent Systems, 2nd ed. CGA G-5.5.
Edeskuty, F.J. and Stewart, W.F. (1996) Safety in the Handling of Cryogenic
Fluids. Plenum Publishing Corporation, New York, Chapter 7.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
692
F. L. Dryer et al.
Erez, L., Sadot, O., Oron, D., Erez, G., Levin, L.A., Sharvts, D., and Ben-Dor,
G. (2000) Study of the membrane effect on turbulent mixing measurements
in shock tubes. Shock Waves, 10, 241–251.
Eubank, C.S., Rabinowitz, M.J., Gardiner, W.C. Jr., and Zellner, R.E. (1981)
Shock-Initiated Ignition of Natural Gas-Air Mixtures. Proc. Combust.
Instit., 18, 1767–1774.
Fay, J.A. and Lekawa, E. (1956) Ignition of combustible gases by converging
shock waves. J. Appl. Phys., 27, 261–266.
Fischer, M. (1986) Safety aspects of hydrogen combustion in hydrogen energy
systems. Int. J. Hydrogen Energ., 11, 593–601.
Fuels and Combustion Research Laboratory (2005) [Online]. Available: http://
www.princeton.edu/~combust/ [5 August 2005].
Gelfand, B.E., Komik, S.V., Bartenev, A.M., Medvedev, S.P., Gronig, H, and
Olivier, H. (2000) Detonation and deflagration initiation at the focusing of
shock waves in combustible gas mixtures. Shock Waves, 10, 197–204.
Gregory, D.P. and Pangborn, J.B. (1976) Hydrogen energy. Annu. Rev. Energy, 1,
279–310.
Hord, J. (1978) Is hydrogen a safe fuel? Int. J. Hydrogen Energ., 3, 157–176.
Houf, W.G. and Schefer, R.W. (2004) Determination of separation distances
from large-scale hydrogen releases. International Code Council (ICC) Hydrogen Ad Hoc Committee Meeting, Coral Gables, FL.
Huang, J, Hill, P.G., Bushe, W.K., and Munshi, S.R. (2004) Shock-tube study of
methane ignition under engine-relevant conditions: experiments and modeling. Combust. Flame, 136, 25–42.
INGAA Foundation (2004) An updated assessment of pipeline and storage infrastructure for the North American gas market: adverse consequences of delays
in the construction of natural gas infrastructure. Report F-2004-01, Energy
and Environmental Analysis, Inc., Arlington, VA.
Jiang, Z., Takayama, K., Babinsky, H., and Meguro, T. (1997) Transient shock
wave flows in tubes with a sudden change in cross section. Shock Waves,
7, 151–162.
Kaltayev, A., Leblanc, J.E., and Fujiwara, T. (1999) Influence of turbulence on
the deflagration to detonation transition in a tube. Proceedings from the
17th International Colloquium on the Dynamics of Explosions and Reactive
Systems, Heidelberg, Germany.
Kruse, B., Grinna, S., and Buch, C. (2002) Hydrogen-Status and Possibilities.
[Online]. Available: http://www.bellona.no/data/f/0/26/97/0 9811 1/Hydrogen
6-2002.pdf [15 May 2005].
Lewis, B. and von Elbe, G. (1987) Combustion Flames and Explosions of Gases.
Academic Press, New York.
Li, J., Zhao, Z., Kazakov, A., and Dryer, F.L. (2004) An updated comprehensive
kinetic model of hydrogen combustion. Int. J. Chem. Kinet., 36, 566–575.
Liñán, A. and Crespo, A. (1976) An asymptotic analysis of unsteady diffusion
flames for large activation energies. Combust. Sci. Tech., 14, 95–117.
Liu, Y.-F., Sato, H., Tsuboi, N., Hjigashino, F., and Hayashi, A.K. (2006)
Numerical simulation on hydrogen fuel jetting from high pressure tank.
Sci. Tech. Energ. Mater., 67, 7–11.
Downloaded By: [Princeton University] At: 19:20 15 March 2007
Spontaneous Ignition of Hydrogen and Natural Gas
693
Lutz, A.E., Kee, R.J., and Miller, J.A. (1987) SENKIN: a Fortran program for
predicting homogeneous gas phase chemical kinetics with sensitivity analysis,
Report SAND87-8248, Sandia National Laboratories, Albuquerque, NM.
National Hydrogen Association (2005) Hydrogen fact sheet 1.008. [Online].
Available: http://www.hydrogenassociation.org/H2-Safety-Fact-Sheet.pdf
[l3 April 2005].
National Hydrogen and Fuel Cell Codes and Standards Coordinating Committee
(2005). [Online] Available: http://www.hydrogenandfuelcellsafety.info/ [20
September 2005].
National Research Council (1995) Prudent Practices in the Laboratory: Handling
and Disposal of Chemicals. National Academy Press, Washington, DC,
p. 123.
NaturalGas.org. (2004) The Transportation of Natural Gas. [Online]. Available:
http://www.natural gas.org/naturalgas/transport.asp [14 July 2005].
NFPA (2005) NFPA 55: Standard for Storage, Use, and Handling of Compressed
Gases and Cryogenic Fluids in Portable and Stationary Containers, Cylinders,
and Tanks. National Fire Protection Association, Quincy, MA.
NFPA. Hydrogen Safety Symposium 2005. NFPA World Safety Conference &
Exposition, Mandalay Bay Resort & Convention Center Las Vegas, NV,
June 6–7, 2005.
Rifkin, J. (2002) The Hydrogen Economy, Penguin Putnam Inc., New York.
Schefer, R.W., Houf, W.G., Bourne, B., and Colton, J. (2004) Turbulent hydrogen-jet flame characterization. Int. J. Hydrogen Energ., submitted.
Schefer, R.W., Houf, W.G., Moen, C.D., Chan, J.P., Maness, M.A., Keller, J.O.,
Leon, M.V. and Tarn, R. (2004) Hydrogen Codes and Standards Unintended Release Workshop. Sandia National Laboratories, Livermore, CA.
Shapiro, A.H. (1954) The Dynamics and Thermodynamics of Compressible Fluid
Flow. The Ronald Press Company, New York.
Shinnar, R. (2003) The hydrogen economy, fuel cells and electric cars. Technol.
Soc., 25, 455–476.
Sidorov, V.P. and Dorofeev, S.B. (1998) Influence of initial temperature, dilution and
scale on DDT conditions in hydrogen-air mixtures. Arch. Combust., 18, 87–103.
Smith, G.P., Golden, D.M., Frenklach, M., Eiteneer, B., Goldenberg, M.,
Bowman, C.T., Hanson, R.K., Song, S., Gardiner Jr., W.C., Lissianski,
V.V., and Qin, Z. (1999) GPJ-MECH 3.0. [Online]. Available: http://www.
me.berkeley.edu/gri_mech/ [2 February 2005].
Swain, M. (2004) Hydrogen Properties Testing and Verification. Hydrogen and
Fuel Cells Summit VIII, Miami, FL.
Taylor, G.I. (1950) The instability of liquid surfaces when accelerated in a direction perpendicular to their planes. Proc. R. Soc. A, 201, 192–196.
Thomas, E. (1997) Direct-Hydrogen-Fueled Proton-Exchange-Membrane Fuel
Cell System for Transportation Applications: Hydrogen Vehicle Safety
Report. DOE report DOE=CE=50389-502, Ford Motor Co., Dearborn, MI.
Transport Canada Prairie and Northern Region (2003) Arctic gas pipeline
construction impacts on northern transportation. [Online]. Available: http://
www.arcticgaspipeline.com/Reference/Documents&Presentations/Canada/
l-03ArcticGasPipelineConstructionSTUDY.pdf [21 June 2005].
Downloaded By: [Princeton University] At: 19:20 15 March 2007
694
F. L. Dryer et al.
Vidicom Media Productions (2005) The Hindenberg. [Online]. Available: http://
www.vidicom-tv.com/hindenburg/index.htm [18 May 2005].
Wang, B.L. and Rehm, W. (2001) Numerical studies of shock reflection and
ignition. Proceedings from the 18th International Colloquium on the Dynamics
of Explosions and Reactive Systems, Seattle, WA.
Wolanski, P. and Wojcicki, S. (1972) Investigation into the Mechanism of the
Diffusion Ignition of a Combustible Gas Flowing into an Oxidizing Atmosphere. Proc. Combust. Instit., 14, 1217–1223.
Zuttel, A. (2003) Materials for hydrogen storage. Mater. Today, 6, 24–33.